BUS603: Managing People

Table 1 presents the team-level correlation matrix, the intraclass correlation estimates [ICC(1) and ICC(2)] for group potency at each time point, and Cronbach's alpha internal consistency estimates. Notably, the ICC estimates increased slightly over time, indicating a growing proportion of variance in group potency that could be attributed to the team-level rather than the individual-level. This suggests increasing consensus in perceptions of team potency over time and stronger emergence. We revisit this pattern to more formally substantiate the emergence of group potency and provide a test of Hypothesis 4.

Table 2 presents the results of the longitudinal measurement invariance analyses. The configural invariance model demonstrated adequate fit, CFI = 0.95 and RMSEA = 0.06. Adding equality constraints on the factor loadings resulted in ΔCFI = –0.001 and ΔRMSEA = –0.002, supporting metric invariance. This suggests that the potency measure retains a similar meaning across occasions. The scalar invariance model resulted in a ΔCFI = –0.003 and ΔRMSEA < 0.0004 versus the metric invariance model. This lends support to scalar invariance, which suggests that the potency measure functions similarly over time. As a final stage in the invariance analyses, additional equality constraints were placed on respective item residuals to assess strict invariance. This model resulted in ΔCFI = 0.003 and ΔRMSEA = –0.004, supporting strict invariance, and suggests that each item had equivalent reliability over time. Together, these invariance analyses suggest equivalence of group potency over time, facilitating our focal latent growth models.

Given the ICCs provided support for aggregating group potency to the team-level, we averaged individual members' group potency scores within each team, and used the aggregated scores to estimate our latent growth model. The unconditional growth model demonstrated adequate fit to the data, χ²(1) = 0.73, p = 0.39, CFI = 1.00, RMSEA = 0.00. With respect to Hypothesis 1, the mean of the latent slope was of central interest, which was estimated as −0.07, p < 0.05. This supported Hypothesis 1, suggesting that group potency decreased over time (by 0.07 units at each time point). The estimate of the latent intercept was 4.06, p < 0.01, and the variances for the latent intercept and slope were 0.20, p < 0.01, and 0.04, p < 0.05, respectively. The correlation between the latent intercept and slope was −0.14, p = 0.59. Interestingly, freeing the slope's factor loading for the second group potency measure, as in a latent basis model did not suggest an improvement in fit. Specifically, Δχ²(1) = 0.71, p = 0.39, and both the Akaike Information Criteria and Bayesian Information Criteria were higher in the latent basis model than the latent growth model. Thus, based on parsimony, we proceed with the linear latent growth model. Notably, even in the latent basis model, the trend did not deviate significantly from a linear trajectory, thus lending further credibility to Hypothesis 1, and the underlying linear, downward pattern of change in group potency. Next, incorporating team effectiveness as a simultaneous outcome of the latent intercept and slope factors also resulted in adequate model-data fit: χ² (2) = 1.22, p = 0.54, CFI = 1.00, RMSEA = 0.00. Specifying regressions between both intercept and slope factors and effectiveness revealed that the regression of effectiveness on the latent slope was b = 0.07, p = 0.99, but that for the latent intercept it was b = 10.23, p < 0.01. Thus, there was no influence of change in potency on team effectiveness, but the starting point of teams' potency was positively related to effectiveness. Accordingly, Hypothesis 2 was not supported, whereas Hypothesis 3 was supported.

To more formally assess the emergence of the group potency construct, we used Lang et al. consensus emergence model. This model uses longitudinal changes in the individual-level residual variances as evidence of emerging consensus. Specifically, decreasing residual variances can be taken as indicative of increasing consensus emergence, and therefore reflects more agreement about a team-level phenomenon. Indeed, in our model the estimated change in residual variance was δ = –0.11, p < 0.05. This suggests significantly less individual-level variance and comparably greater sharedness at the team-level over time. In other words, this negative coefficient supports the proposition that group potency demonstrated significant increases in the support for emergence over the three measurement occasions. Thus, Hypothesis 4 was supported.

Finally, we incorporated the conscientiousness and extraversion team-level predictors into the latent growth model. This also resulted acceptable model-data fit: χ² (4) = 3.02, p = 0.55, CFI = 1.00, RMSEA = 0.00. Neither of the indirect effects involving the latent slope had 95% CIs that excluded zero: the conscientiousness → latent slope → team effectiveness indirect effect was −0.01, 95% CI = –3.11–2.65, and the extraversion → latent slope → team effectiveness indirect effect was 0.08, 95% CI = –2.36–4.55. The indirect effect involving extraversion → latent intercept → team effectiveness was also not significant, −0.73, 95% CI = –9.34–4.65. However, the indirect effect of conscientiousness → latent intercept → effectiveness was significant, 5.57, 95% CI = 0.59–23.58. Thus, there was no evidence for the mediating role for change in potency, but instead the latent intercept transmitted the effect of conscientiousness on team effectiveness. In sum, Hypothesis 5a was supported, but Hypotheses 5b, 6a, and 6b did not receive support.